Looks amazingly natural. Any chance of a DMD2 version of it?

Finally, an XL checkpoint that's not the same as every other XL out there, and it truly can generate amazing outputs extremely quickly.

...except.. we don't need more speed for XL. The thing this actually needs, is a NOT DMD2 version, one with normal CFG and most importantly, a negative prompt and LORA compatibility. Unfortunately, I have a feeling that's not coming (feel free to prove me wrong).

As it is, this checkpoint is nothing more but a neat experiment. I mean, it's nice for playing around, it's just unusable in any serious workflow.

Yes! Celeb and lora character compability is the first priority now. We have good hardcore models already, they just need to be compatible.

I tried merge Lustify with an all-rounder model for pure celeb and lora adherence, result was shit, not good.

Maybe you can make one.

Edit: Op is awesome, and added some gallery examples showing the model does well with generating men! I’m going to give it a try with some more gay sexual situations and report back, too.

If you’re going to use such boldly gpt-written marketing jargon as “Sophisticated Artistic Versatility” you should at least have one sample image showing that the model can generate images of gay men engaged in “themes of intimacy.”

It won’t make you any less hetero, but it will help those of us who aren’t to know if your model is usable.

I’d even settle for just one image of a fully clothed man without a woman, just to prove that the model can do basic anatomy well enough.

you got some bigasp2 in this model ?

all that prompt is a bit long isn't sdxl text encoder can carry tokens like 77 only ?